Deforestation presents significant challenges for land-use mon itoring, requiring advanced computational approaches to derive mean ingful insights from spatio-temporal data. This paper introduces a novel interest-driven pattern mining framework to analyze deforestation dy namics in the Kroumiria region of Tunisia from 2001 to 2022. By lever aging frequent pattern mining techniques, specifically the FP-Growth al gorithm, and validating the results through statistical permutation tests, our methodology ensures both interpretability and reliability. Unlike tra ditional GIS-based or predictive models, this approach prioritizes pat terns exhibiting statistically significant variations before and after the 2011 revolution. Our findings highlight the effectiveness of integrating machine learning-based pattern mining with statistical validation to ex tract actionable knowledge from large-scale environmental datasets, ul timately enhancing decision-making processes in land management and AI-driven spatial analytics |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.